WorkplaceAI Act
AI images on screen: when disclosure is required
Article 50 of the EU AI Act has applied since 2 August 2026. For most creatives little changes; for deceptively realistic depictions, a lot.

An image for a screen is often produced in seconds today: one sentence typed into a program, and the winter landscape for the offer of the week is ready. Since 2 August 2026, Article 50 of the EU AI Act has applied to this. It governs when artificially generated content has to be marked or disclosed. For editorial teams, marketing departments and agencies supplying screens it deserves a close look, because the duties fall on two different roles.
Two roles: provider and deployer
The regulation distinguishes the provider of an AI system, meaning the maker of the image generator, from the deployer who uses it. The deployer is the agency or department that has the image generated. The term is not the same as the operator of a screen, although the two may coincide.
What the provider must do
Under paragraph 2, providers of systems that generate synthetic audio, image, video or text content must ensure that the outputs are marked in a machine-readable format and detectable as artificially generated. This is a technical marker in the file, not a visible caption.
Systems that merely perform an assistive function for standard editing, or do not substantially alter the input, are exempt. Cropping or brightening a photograph does not create synthetic content.
What the deployer must disclose
Paragraph 4 addresses those who use the tool. Anyone who generates or manipulates image, audio or video content constituting a deep fake must disclose that the content has been artificially generated or manipulated.
The definition is what matters. Under Article 3, a deep fake is AI-generated or manipulated content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful. Two conditions therefore have to be met together: resemblance to something real, and the false impression of authenticity.
A recognisably drawn illustration does not meet them. A photorealistic image of one's own branch with a queue outside the door that never existed may well do so. So may a video in which a real person says something they never said.
For evidently artistic, creative, satirical or fictional works there is a relief: disclosure then only has to be made in a manner that does not hamper the display of the work.
A second duty concerns text published to inform the public on matters of public interest. It does not apply where the text has been reviewed by a human and someone holds editorial responsibility.
When the information has to be given
Under paragraph 5 the information must be provided at the latest at the time of the first interaction or exposure, clearly and distinguishably. On a screen that means the notice belongs in the creative itself, not on a website the viewer never sees.
The December deadline
For providers the legislator made an adjustment in summer 2026. The amending Regulation (EU) 2026/1744 grants systems that were on the market before 2 August 2026 a transitional period: their providers must implement machine-readable marking by 2 December 2026. There is no such period for the deployers' disclosure duty.
The Commission has published a code of practice on marking and labelling. Adherence is voluntary; the obligations under Article 50 are not.
Three rules for everyday work
- Record the origin. For every creative, note whether and with what it was generated. Without that record nothing can be checked later.
- Label what is deceptively real. If a generated image looks like a photograph of something real, a visible notice goes into the creative.
- Do not strip the marker. The provider's machine-readable marking should survive export and conversion.
This article reflects the text of the regulation and is not legal advice.


